[Harzva]

Change

Readers had to leave a scenario card to find its practice, search landed at the whole card collection, and copying practice materials omitted the actual task. Version 0.9.3 adds direct market/card links, individual scenario search targets, and complete first-role tasks with fictional inputs. Task, teaching-reference and revision blocks have distinct bilingual headings and copy labels; copied code remains exact.

The content pass clarifies role/workflow selection, catalogue counts, Agent/host/model terms, append-only import records and recovery after accidentally sharing sensitive material. Missing host IDs remain insufficient evidence for an upgrade. Narrow CSS changes preserve the existing reading style while improving heading spacing, a small brand descriptor and count contrast.

Evidence

• Primary sources: existing handbook, scenario and configuration contracts; official-source recheck documented in docs/handbook-sources-20261025.md.
• Original AI-assisted project content under the repository license. Actual Kimi CLI consultation completed; unsupported identity/deadline suggestions were rejected and credential/data handling corrected.
• Independent bilingual content review and native CX Luna visual review completed. No real-account Bot execution is claimed; examples remain fictional teaching material.
• Actual host import, accounts and tool authorization remain separate from website reading and downloads.

Checks

• Stable Bot IDs, role/workflow counts and published source versions retained; no new Bot entries.
• No credentials, private memory, conversations or unlicensed prompts added.
• Build, catalogue check, 69 tests, site generation and public audit passed.
• 1,125 protected entry/template/registry/localization/scenario source files match 0.9.2.
• All 30 labeled practice copies retain exact text, with a verified denied-clipboard fallback.
• Ten localized market links, individual scenario search, bilingual anchors, Markdown, no-JavaScript reading and 390/1440px page bounds verified.
• Final visual follow-up: heading spacing, 11px desktop descriptor, 16.83:1 count contrast and 375/390px header bounds checked.